Stan,
I always love your replies. I may often NOT comprehend them, and, reply
badly, but I do read them and continue to think about the subject.
I never meant to start a "skirmish" over RAM.
Yes, I did buy/own DDR3 RAM for my 3 P5Q3 m/b's. As I read the notes, it
was a required upgrade. OK. I am an early adopter. So far, so good. I
have no complaints yet about my "not much more expensive that DDR2" RAM.
Here, it is all good ATM. But then, I do NOT exercise my RAM to the
limits I think both you and James are sharing.
I am mostly pedestrian in the RAM-speed department now.
Best,
Duncan
Stan Zaske wrote:
I understand that i7 has no DDR2 option as that has been widely reported
for many months. I thought that I was pretty clear in saying there is
little to no performance difference in going from one memory standard to
the next and therefore if you choose to buy into any other mainstream
(non-i7) platform there is no practical reason to buy DDR3 over DDR2
until there is no other choice.
i7 of course is a fine platform and the fastest there is currently (as
everyone agrees) but thats not because of DDR3 thats because of the i7's
architecture. In fact, there have been several reviews comparing i7 dual
channel vs. triple channel and there isn't any difference again even
though the synthetic benches say there is.
DDR3 is currently too slow and has too high a latency to demonstrate a
clear performance advantage over DDR2 and it was the same exact
situation during the transition from DDR to DDR2 several years ago as
I'm sure you must remember. It was severely hyped and when the
benchmarking began it was a huge disappointment to hardware enthusiasts
hoping for something great. I know I was disappointed after all the BS.
James Boswell wrote:
Well, the issue there is that I can't benchmark an i7 with DDR2
and as I've said already, due to the FSB bottleneck, it really is
completely useless on a Core 2
On 2 Jun 2009, at 01:30, Stan Zaske wrote:
Can you get me a link to a good hardware review that show these real
world performance improvements that you're talking about? As I said,
I've been reading hardware reviews for many years on this subject and
see no hard evidence to back what you're saying. Synthetic benchmarks
always show an improvement in bandwidth but when you use practical
software applications and do comparisons there is minimal if any
difference in speed. Certainly nothing that a human beings perception
could detect.
James Boswell wrote:
On 2 Jun 2009, at 01:00, Stan Zaske wrote:
SDRAM--->DDR--->DDR2--->DDR3=more bandwidth, more latency, lower
voltage and lower heat. After a decade and dozens if not hundreds
of hardware reviews that show minimal to no real world performance
improvements (synthetic? Uh huh!) DDR3 is another solution to a
problem that doesn't exist. Only when speeds get well above DDR3
2500 will there be significant and noticeable speed improvement.
DDR3-1600 is a very significant improvement over DDR2-800... the
issue is that Core 2's are hooked up to memory controllers the other
side of the an FSB, the IMC in the i7 delivers... heroic bandwidth
from DDR3
The problem does exist, but the memory always always ALWAYS comes
before its needed, in this case it came in the middle of the Core 2
generation, when it wasn't needed or useful until i7
One exception to this is DDR memory coming with the Athlon (huge
performance kick)
-JB
